Modelamiento y selección de procesos productivos

2.2.1

Study Subjects and Logistics

All subjects provided written informed consent to a study protocol approved by a local research ethics board and Health Canada. Subjects were enrolled between the ages of 60 and 90 years with a smoking history of < 0.5 pack-years and no history of chronic respiratory or cardiovascular disease. After providing consent, all subjects made a single 2-3 hour visit and underwent the following evaluations in the same order: 1) Burden of Obstructive Lung Disease (BOLD) occupational questionnaire,5 2) spirometry, 3) plethysmography and the diffusing capacity of carbon monoxide (DLCO), 4) CPET

including dyspnea score, and, 5) MRI. Pulmonary function tests were completed in approximately 45 minutes followed by CPET that was completed in 10-15 minutes. MRI was performed following CPET and for subjects who did not have ventilation defects and those with defects that responded to DI, MRI was completed within 10-15 minutes. For subjects who were administered salbutamol, MRI was performed 25-30 minutes post salbutamol inhalation.

2.2.2

Questionnaires

All subjects completed the BOLD occupational questionnaire21 with exposures defined as previously described22 including: 1) organic dust (farming; flour, feed or grain milling and cotton or jute processing), 2) inorganic dust (working with asbestos; hard-rock mining; coal mining; sandblasting and foundry or steel milling), and, 3) irritant gases (welding; firefighting; chemical or plastics manufacturing). We also directly queried subjects about spousal and/or life-partner smoking history and recorded potential household 2nd hand- smoke exposure.

63

2.2.3

Pulmonary Function and Cardiopulmonary Exercise Tests

Spirometry was performed using an ndd EasyOne spirometer (ndd Medizintechnik, Zurich) according to American Thoracic Society (ATS) guidelines. Body plethysmography was performed for the measurement of lung volumes and DLCO was measured using the gas

analyser (Medgraphics Corporation, St. Paul, MN). Inspiratory capacity (IC), defined as the volume change recorded at the mouth when taking a slow full inspiration from a position of passive end-expiration, was measured using body plethysmography31 as was airways resistance (Raw), defined as the pressure difference per unit flow. Post-DI pulmonary function tests were not performed but for subjects administered salbutamol, FEV1 and FVC were recorded 35-40 minutes post-inhalation and upon completion of MRI.

After completion of pulmonary function tests, all subjects performed CPET using a cycle ergometer (MedGraphics Ultima PFX, MedGraphics Corporation) according to ATS guidelines 14 with a two minute warm-up with no resistance, followed by a 20 watt incremental increase in work rate. Subjects were required to pedal at a frequency of 60 rpm with increasing work rate until the ventilatory anaerobic threshold was reached. Ventilatory anaerobic threshold was defined as the time when CO2 production increased

disproportionately in relation to O2 consumption. The ventilatory anaerobic threshold was

determined by onboard software for the CPET unit which used an iterative regression and analysis of slope. Subjects continued exercise until fatigue was reported. Oxygen uptake (VO2), respiratory exchange ratio (RER), work rate (i.e. power) as well as minute

ventilation (VE) were measured at rest, at the ventilatory anaerobic threshold, and when

maximum pulmonary O2 uptake was reached. The time taken to reach VO2max was also

recorded. CPET measurements including VO2, power, and VE were adjusted for age, sex,

and height using percent predicted values previously reported.17,18,30 _{Borg dyspnea and leg}

discomfort scales were used before and after exercise.

2.2.4

Image Acquisition

MRI was performed after completion of all pulmonary function and cardiopulmonary exercise tests on a 3.0 Tesla MR750 (General Electric Health Care (GEHC), Milwaukee, WI) system as previously described23 using a whole-body gradient set with maximum

64

gradient amplitude of 1.94 G/cm and a single-channel rigid elliptical transmit/receive chest coil (Rapid Biomedical GmbH, Germany). For both 1H and 3He MRI, subjects were instructed to inhale a gas mixture from a 1.0-liter Tedlar bag (Jensen Inert Products, Coral Springs, FL) from FRC, and image acquisition was performed during a 16 second breath- hold. Coronal (anatomical) 1_{H MRI was performed using the whole-body RF coil and} 1_H

fast-spoiled, gradient-recalled echo sequence using a partial echo (16s total data acquisition, repetition time (TR)/echo time (TE)/flip angle = 4.7ms/1.2ms/30°, field-of- view (FOV)=40×40cm, bandwidth=24.4 kHz, matrix=128×80, 15-17 slices, 15mm slice thickness, 0 gap), as previously described.23 3He MRI static ventilation images were acquired using a fast gradient-echo method using a partial echo (14s total data acquisition; TR/TE/flip angle = 4.3ms/1.4ms/7°, FOV = 40×40cm, bandwidth=48.8 kHz, matrix= 128×80, 15-17 slices, 15mm slice thickness, 0 gap).

Immediately after image acquisition, at the scanner while the subject was still in position,

3_{He static ventilation images were qualitatively evaluated for ventilation abnormalities by}

a single trained observer. If 3He gas was homogeneously distributed throughout the lung and there were no visible ventilation defects, the subjects were classified as belonging to the no defect group and the session was deemed complete. In contrast, if 3He gas was heterogeneously distributed throughout the lung and/or there were visually obvious ventilation defects, the subject was classified as having ventilation defect(s). Upon qualitative inspection, subjects with visually obvious ventilation defects were instructed to perform DI. They were instructed to sit up whilst remaining on the scanner bed and take four deep breaths in through their nose and out through their mouth. Imaging was performed immediately following DI. If defects persisted following DI, the subject inhaled four puffs (400 µg) of salbutamol while seated upright, and 25 minutes later imaging was performed on a final occasion.

2.2.5

Image Analysis

Based on visual inspection at the scanner, subjects were classified as: 1) no defects—those subjects without visually obvious ventilation defects, 2) subjects with ventilation defects that responded to DI or salbutamol, and, 3) subjects with ventilation defects that did not

65

respond to DI or salbutamol. 3He MRI semi-automated segmentation was performed, as previously described,31 to generate ventilation defect percent (VDP) – the ventilation defect volume (VDV) normalized to 1H MRI thoracic cavity volume (TCV). Briefly, 3He static ventilation images were segmented using a k-means approach that classified voxel intensity values into five clusters ranging from signal void (cluster 1[C1] or VDV) and hypo-intense (cluster 2 [C2]) to hyper-intense signal (cluster 5 [C5]), therefore, generating a gas distribution cluster-map. For delineation of the ventilation defect boundaries, a seeded region-growing algorithm was used to segment the 1_{H MRI thoracic cavity for registration}

to the cluster-map, as previously described.31

Ventilation heterogeneity was estimated according to a previously described method27 using the coefficient of variation (COV). A local COV, rather than a global COV, was generated27 to ensure local ventilation heterogeneity was not ignored. Briefly, for each voxel in a region of interest (ROI) a local ventilation heterogeneity value was calculated by computing the COV of the signal intensity in a 5×5 voxel neighbourhood which corresponded to a 244 mm2 area centred on that voxel. To ensure that the 5×5 voxel neighbourhood did not include voxels that were outside of the lungs, a signal-to-noise threshold of five was established. For example, any voxels that were in a neighbourhood with an overall signal-to-noise ratio of less than five were excluded from the COV computation. COV of the lung was calculated for each subject as in Equation 2-1:

Equation 2-1 𝐶𝑂𝑉 = √∑ 𝜎𝑖2 𝑁 𝑖=𝑁 𝑖=1 ∑𝑖=𝑁𝜇_𝑖 𝑖=1 𝑁

where 𝜎𝑖is the standard deviation of the signal intensity of the 5×5 voxel neighbourhood,

𝜇_𝑖 is the mean signal intensity of the 5×5 voxel neighbourhood, and 𝑁is the number of ventilation ROIs in the lung. The ventilated lung ROI was defined as gas distribution cluster-map clusters C2-C5.

66

We also measured regional differences in VDP for the centre nine 15mm thick slices using several different measurements as shown in Figure 2-1. The VDP gradient in the posterior- anterior direction was defined as the slope of the line that described the change in VDP from the most posterior slice in centimeters over the nine central slices (with a slice thickness of 15mm each). The centre nine slices were used to ensure the static ventilation slices across subjects had adequate signal-to-noise ratios (i.e. SNR>10) for VDP calculations as well as to maintain an equal number of slices among all subjects. To calculate the gradient in the superior-inferior (SI) direction, the centre nine coronal slices were reformatted into 15mm axial slices. The VDP SI gradient was defined as the slope of the line that described the change in VDP over the axial superior-inferior slices. In addition, the VDP difference between the most posterior and anterior slices (VDP ∆PA) was calculated as was the VDP difference between the most superior and inferior slices (VDP ∆SI) of the central nine slices. The ventilation defect percent located on the peripheral boundary of the lung (relative peripheral VDP) was estimated as the ratio of the ventilation defect volume for the outermost 10 voxels of each slice (not including boundary voxels defined as SNR <2) to the 1H MRI thoracic cavity volume. The proportion of ventilation defects located on the peripheral boundary was quantified as the ratio of peripheral lung VDV to whole lung VDV (i.e. VDVPer/VDVWL).

Figure 2-1 Schematic for regional evaluation of 3He MRI ventilation-defect percent (VDP) The central coronal static ventilation image slices (9 × 15 mm) were evaluated. The lung was also divided into 15-mm slices in the axial direction. The VDP ΔSI was defined as the

67

difference between the most superior slice and the most inferior slice. The VDP ΔPA was defined as the difference between the most posterior slice and the most anterior slice.

2.2.6

Statistical Analysis

Analysis of variance (ANOVA), multivariate regression and post-hoc analysis using the Holm Bonferroni correction 32 were performed using SPSS 20.0 (IBM, Armonk, NY). Paired sample t-tests were performed to determine the differences in VDP ΔPA and VDP ΔSI using SPSS 20.0. We used the NHANES III reference standards33 _{for percent}

predicted values. Univariate relationships were determined using regression (r2) and Pearson correlation coefficients (r) for all subjects with GraphPad Prism V.6.00 (GraphPad Software Inc, La Jolla, CA). Stepwise multivariate regression was used to identify the predictors of VDP. The variables considered for modelling were chosen based on statistically significant univariate relationships with VDP. Results were considered statistically significant when the probability of making a Type I error was less than 5% (p<0.05).